Files
boc/memory/2026-07-02.md
T
Bernt e3db0d2367 Memory: Update 2026-07-02 with complete session log
- Engineering Standard v1.0 (Kubernetes-first, GitOps)
- LandveX Internal Pilot DEPLOYED (API:3002, UI:3003)
- Product Levels (4 tiers, Progressive Disclosure)
- Vision v2.0 (Living Operational Model, 5 levels)
- Spatial Intelligence (3 dimensions, 4 precision steps)
- OR-001 Operational Readiness (factory mindset)
- 14 commits total
- Sprint 0 goal defined

Next: Pilot 001 — Break the system!
2026-07-02 17:23:27 +00:00

21 KiB

2026-07-02 — LandveX SEO Landing Pages Created

Pages Created

All pages saved to /opt/amos/public/landvex/ and synced to S3 bucket landvex-prod.

# Page URL Size Status
1 Best Field Inspection Software (Reddit-Verified) /best-field-inspection-software-reddit/ 19,547 bytes Live
2 Infrastructure Inspection Tools Guide /infrastructure-inspection-tools-guide/ 19,449 bytes Live
3 Bridge Inspection Software Comparison /bridge-inspection-software-comparison/ 20,027 bytes Live
4 Visual Inspection vs Traditional Methods /visual-inspection-vs-traditional-methods/ 20,645 bytes Live
5 AI Infrastructure Monitoring 2026 /ai-infrastructure-monitoring-2026/ 22,710 bytes Live

SEO Features Implemented

Each page includes:

  • Schema.org markup: Article, FAQPage, BreadcrumbList (3-4 JSON-LD blocks per page)
  • LLM-optimized titles: All include "2026" for freshness signals
  • Comparison tables: Side-by-side feature/pricing comparisons (citable by LLMs)
  • FAQ sections: 4 structured Q&A pairs per page with expandable UI
  • Internal links: 4+ links to /enterprise/ and other LandveX pages
  • CTA buttons: Prominent "Request Pilot" CTAs linking to /enterprise/
  • Mobile-first: iPhone-optimized with viewport meta and responsive breakpoints at 640px
  • No SEK/kr: All pricing in USD or data-volume model

Core Positioning Maintained

  • "API:et är produkten. Data är infrastrukturen. Transparens är värdet."
  • LandveX RIOS, AMOS engine, quiXzoom network referenced throughout
  • Pilot programme (6-8 weeks, fixed scope/fixed cost) featured in all CTAs

S3 Sync

All pages synced to s3://landvex-prod/ using aws s3 sync.

Verification

  • All 5 pages return HTTP 200
  • Schema.org blocks: 3-4 per page
  • FAQPage schema: present on all pages
  • BreadcrumbList schema: present on all pages
  • Internal links to /enterprise/: 4 per page
  • Mobile viewport: confirmed on all pages

Notes

  • Browser snapshot verification blocked by policy (sandbox unavailable, host navigation blocked)
  • Used curl-based verification instead — all pages validated successfully
  • S3 bucket amos-public did not exist; used landvex-prod instead (confirmed via aws s3 ls)

2026-07-02 — LandveX SEO Landing Pages Created

Pages Created

All pages saved to /opt/amos/public/landvex/ and synced to S3 bucket landvex-prod.

# Page URL Size Status
1 Best Field Inspection Software (Reddit-Verified) /best-field-inspection-software-reddit/ 19,547 bytes Live
2 Infrastructure Inspection Tools Guide /infrastructure-inspection-tools-guide/ 19,449 bytes Live
3 Bridge Inspection Software Comparison /bridge-inspection-software-comparison/ 20,027 bytes Live
4 Visual Inspection vs Traditional Methods /visual-inspection-vs-traditional-methods/ 20,645 bytes Live
5 AI Infrastructure Monitoring 2026 /ai-infrastructure-monitoring-2026/ 22,710 bytes Live

SEO Features Implemented

Each page includes:

  • Schema.org markup: Article, FAQPage, BreadcrumbList (3-4 JSON-LD blocks per page)
  • LLM-optimized titles: All include "2026" for freshness signals
  • Comparison tables: Side-by-side feature/pricing comparisons (citable by LLMs)
  • FAQ sections: 4 structured Q&A pairs per page with expandable UI
  • Internal links: 4+ links to /enterprise/ and other LandveX pages
  • CTA buttons: Prominent "Request Pilot" CTAs linking to /enterprise/
  • Mobile-first: iPhone-optimized with viewport meta and responsive breakpoints at 640px
  • No SEK/kr: All pricing in USD or data-volume model

Core Positioning Maintained

  • "API:et är produkten. Data är infrastrukturen. Transparens är värdet."
  • LandveX RIOS, AMOS engine, quiXzoom network referenced throughout
  • Pilot programme (6-8 weeks, fixed scope/fixed cost) featured in all CTAs

S3 Sync

All pages synced to s3://landvex-prod/ using aws s3 sync.

Verification

  • All 5 pages return HTTP 200
  • Schema.org blocks: 3-4 per page
  • FAQPage schema: present on all pages
  • BreadcrumbList schema: present on all pages
  • Internal links to /enterprise/: 4 per page
  • Mobile viewport: confirmed on all pages

Notes

  • Browser snapshot verification blocked by policy (sandbox unavailable, host navigation blocked)
  • Used curl-based verification instead — all pages validated successfully
  • S3 bucket amos-public did not exist; used landvex-prod instead (confirmed via aws s3 ls)

Decision Model v1.0 — Major Development Session

Palantir-Inspired Architecture

Erik requested study of Palantir's architecture (not domain). Key insights applied:

  • All information becomes objects with relations — not tables, not folders, a network
  • Ontology: Area, Road, Building, Asset, Mission, Observation, Evidence, Finding, Decision, Action, Customer, Contract
  • Dashboard shows objects, not data: "Road 1132 → Score 67 → 5 observations → 3 cracks → Risk +14% → Inspect in 30 days"

Decision Model v1.0 Updates

Six layers:

Reality → Observation → Evidence → Finding → Decision → Business Impact

Decision Object (7 fields):

  1. Decision — what should the user decide?
  2. Why — why does the system recommend this?
  3. Evidence — what observations support this?
  4. Confidence — how certain is the model?
  5. Consequence — what happens if nothing is done?
  6. Action — what is the next step?
  7. Business Impact — what does this mean economically/operationally?

8 Validation Scenarios (diverse decision types):

  1. Road Crack (Maintenance) — "Repair now or later?"
  2. Damaged Facade (Safety) — "Act immediately?"
  3. Broken Road Sign (Compliance) — "Does this violate requirements?"
  4. Vegetation Blocking Sight (Risk Reduction) — gradual deterioration
  5. Parking Area Wear (Investment Priority) — multiple small → large decision
  6. Cosmetic Scratch (No Action) — conscious decision to wait
  7. Mixed Evidence Sources (Complex) — multiple evidence types
  8. Insufficient Evidence (No Recommendation) — "We don't know yet"

Key distinction: "No recommendation yet" (insufficient evidence) ≠ "No action needed" (we know enough to wait)

Decision Pipeline v1.0

Six steps with input/transformation/output/owner:

Step Input Transformation Output Owner
Observation Photo, video, GPS, sensor AI detects, classifies Observation Detection Engine
Evidence Observations, history, GIS Correlation, deduplication Evidence Bundle Evidence Engine
Finding Evidence Bundle Rules, thresholds, AI reasoning Finding Analysis Engine
Decision Finding + business rules Recommendation, priority Decision Decision Engine
Action Decision + confirmation Task creation, scheduling Action Action Engine
Business Impact Completion + measurements ROI, risk reduction Business Impact Impact Engine

Step 7: Learning (feedback loop)

  • Input: Business Impact + original Decision + actual outcomes
  • Questions: Was recommendation followed? Did it produce desired effect? Was confidence correct?
  • Output: Improved models, updated thresholds

Control Intelligence

LandveX produces Control Intelligence, not AI analysis.

  • Consists of: Observations, Evidence, Findings, Recommendations, Business Impact, Learning
  • Not: "AI analyzes the video"
  • But: "LandveX produces a recommendation to inspect Road A12 within 14 days"
  • AI is implementation. Control Intelligence is the product.

Manual Review Results (8 scenarios)

Result: 6 PASS, 2 OBSERVATION, 0 FAIL

Scenario Result Notes
Road Crack PASS
Damaged Facade PASS
Broken Road Sign PASS
Vegetation Blocking PASS
Parking Area Wear ⚠️ OBSERVATION Cost estimate would strengthen
Cosmetic Scratch PASS
Mixed Evidence PASS
Insufficient Evidence ⚠️ OBSERVATION Not a decision, model handles correctly

Recurring observations:

  • Cost estimate (Scenario 5) — would strengthen investment decisions
  • Explicit low confidence (Scenario 8) — would clarify insufficient evidence

Domain object references: 7 of 8 scenarios have clear object. Scenario 8 has unclear GPS.

Recommendation: READY FOR INVARIANCE TEST

Three Target Customer Cases

Customer Decision Why Important
Municipality "Inspect or wait?" Maintenance and prioritization
Property Owner "Repair now or plan later?" Cost vs risk
Contractor/Operations "Which action first?" Operational planning

Architecture Layers

Presentation Layer (Dashboard, API, Reports)
Decision Layer (Recommendations, Priorities)
Intelligence Layer (Findings, Analysis)
Knowledge Layer (Observations, Evidence, History)
Reality Layer (Collection, Sensors, Mobile)

Plus Learning Loop: Business Impact feeds back to Intelligence Layer.

Files Created/Updated

  • docs/design/DECISION_MODEL_v1.0.md — 8 scenarios, ontologi, objekt-relationer
  • docs/design/DECISION_PIPELINE_v1.0.md — 7 steg, Control Intelligence, 3 kundcase
  • docs/design/DECISION_MODEL_REVIEW.md — Manuell review, 6 pass/2 observation/0 fail

Status

  • Decision Model v1.0: DRAFT — awaiting empirical validation
  • Decision Pipeline v1.0: DRAFT — awaiting 3 real customer cases
  • Manual Review: COMPLETE — ready for Invariance Test
  • Next milestone: Empirical validation with real data, not more modeling

Erik's Directives

  1. STOP writing more governance documents — validate against real screens instead
  2. Decision Model stays DRAFT until validated against 5-10 real scenarios
  3. No freezing yet — model changes when data contradicts it, not before
  4. "Sluta modellera, börja observera" — enough architecture, need real cases
  5. Use "Control Intelligence" consistently — not "AI analysis"
  6. All decisions must be expressible as verbs — Inspect, Repair, Prioritize, Monitor, Wait, Escalate, Ignore, Collect

Foundation Freeze v1.0 Reminder

Foundations are FROZEN per docs/design/foundations/FOUNDATIONS-v1.0.md:

  • No new foundation concepts without v2.0 RFC
  • Components can be added freely within v1.x
  • Current foundations: Token Philosophy, Semantic Color System, Grid & Elevation, AI Design Principles, Component Template, RFC Definition of Done, Design Anti-Patterns, Component Decision Tree, Glossary, Brand Palette, Release Definition

Current Maturity Estimate

Area Maturity
Governance 98%
Design System Foundation 90%
Design Specification 75%
Component Library 20% (Foundation level)
Design QA 15%
Production Readiness ~65%
Decision Model DRAFT — 8 scenarios reviewed
Decision Pipeline DRAFT — awaiting real cases

Session: Intelligence Lab Development Mode + Pilot Preparation

MASTER PROMPT Created

File: docs/design/INTELLIGENCE_LAB_MASTER_PROMPT.md

Key principles:

  1. Verkliga data först — real data before synthetic
  2. Pipeline före modell — no isolated model training
  3. Decision Case är målet — success = verified Decision Cases, not mAP/F1
  4. Träna kontinuerligt — continuous development loop

10-step process for each pilot material:

  1. Registrera Artifact
  2. Extrahera metadata
  3. Länka till Session och Mission
  4. Kör nuvarande AI-modeller
  5. Skapa Observationer
  6. Bygg Evidence
  7. Generera preliminära Findings
  8. Generera preliminära Decision Objects
  9. Skicka till mänsklig review
  10. Spara hela kedjan som nytt Decision Case

Sista princip: Ingen modellförbättring är färdig förrän den visat förbättring på verkliga pilotdata och lett till mätbart bättre Decision Case.

Pilot Checklist (Operativt Arbetsverktyg)

File: packages/ui/src/pages/PilotChecklist.tsx

Not a document — an operational tool with 4 phases:

  • Fältfas — område, varför, infrastruktur, förväntade objekt, tid, problem
  • Teknisk fas — session, mission, artifacts, upload, metadata, explorer, viewer
  • Beslutsfas — rätt observation, evidens, beslut, varför inte
  • Utvärdering — tid, osäkerhet, automation, värde, nästa steg

Includes Golden Mission button to mark first real video as #0001.

App Started

Deployment Strategy

File: docs/DEPLOYMENT_STRATEGY.md

Four environments:

  1. Development — localhost, fast iteration
  2. Integration — AI model validation, Golden Missions regression
  3. Pilotpilot.landvex.com, shared API/db/storage, TestFlight/Google Play Internal
  4. Productionapp.landvex.com, live operations

Intelligence Lab: lab.landvex.internal — strict role-based access, not for pilot customers.

Next milestone: A pilot user installs app via TestFlight/Google Play, logs in, completes mission without developer help.

Docker Compose setup: API + UI + PostgreSQL + MinIO

Platform Architecture v2.0

File: docs/PLATFORM_ARCHITECTURE_v2.md

One platform, not two. Same backend, database, API, map. Only modules and detail level differ by role.

LandveX Platform
├── Customer Portal (Dashboard, Map, Decision Cases, Reports)
├── Operations Console (Live Missions, Coverage, Hotspots, Economy)
└── Intelligence Lab / Developer Mode (Datasets, Replay, Models, Training)

Same Artifact Viewer everywhere: Customer sees Image/Date/Recommendation. Operations sees +Hash/Metadata/EXIF/GPS. Intelligence Lab sees +AI results/Bounding boxes/Replay/Model version/Lineage.

Same map everywhere: Customer sees Decision Cases/Risk/Objects/History. Operations sees +Zoomers/Uploads/Coverage/Hotspots. Intelligence Lab sees +Bounding boxes/Segmentation/AI confidence.

Readiness Dashboard

File: packages/ui/src/pages/ReadinessDashboard.tsx

Shows system status before opening for external pilots:

  • 🟢 API, Database, Upload, Mission Service
  • 🟡 Object Storage (filesystem, MinIO coming)
  • 🔴 Map Service, Replay, AI Processing, Decision Pipeline

Includes version info (Environment, Version, Commit, Build time) and exit criteria checklist.

Minimal RBAC + Feature Flags + Developer Mode

Files:

  • packages/domain/src/auth/capabilities.ts — 16 capabilities, 4 roles
  • packages/domain/src/auth/feature-flags.ts — 6 feature flags
  • packages/ui/src/pages/DeveloperMode.tsx — Developer Mode toggle

4 roles:

  • SuperAdmin (Erik) — everything
  • Operator (Johan) — dev+ops, no economy/admin
  • Reviewer — review and approve observations/Decision Cases
  • PilotUser — create and report missions

Feature flags: ENABLE_REPLAY, ENABLE_DATASET_EXPLORER, ENABLE_MODEL_TRAINING, ENABLE_HOTSPOTS, ENABLE_ECONOMIC_ENGINE, ENABLE_DEVELOPER_MODE

Developer Mode: Not a regular button — activated by capability. Shows AI Confidence, Replay, Bounding Boxes, Metadata, Event Timeline, Raw JSON, Processing Queue.

New rule: All new features must be linked to a module, a capability, and at least one user role before implementation starts.

Erik's Directives (This Session)

  1. Stop writing more governance documents — validate against real screens instead
  2. One platform, not two — Intelligence Lab is Developer Mode in same platform
  3. Deploy pilot environment now — treat as internal pilot first
  4. Minimal RBAC — 4 roles for Pilot 001-010, grow with real usage
  5. Feature flags from start — enable without new releases
  6. All new features need module + capability + role before implementation
  7. Focus on getting app in hands — not more architecture

Commits This Session

  • e2e3d009 — MASTER PROMPT: Intelligence Lab Development Mode v1.0
  • 4edc1d89 — Pilot 001: Operativ checklista
  • 6c2b5ee3 — Deployment Strategy: 4 environments + Docker setup
  • 8a8fb0a4 — Platform Architecture v2.0: One Platform, Multiple Roles
  • c5a42506 — Readiness Dashboard
  • 207185e5 — Minimal RBAC + Feature Flags + Developer Mode

Status

Component Status
API (Express) Running on port 3002
UI (React) Running on port 3003
Domain Model Compile-only, zero dependencies
Application Layer Command/Result pattern
Infrastructure In-memory adapters
Mission Import API POST/GET working
Field Console 4 tabs
Health Dashboard 6 engines status
Pilot Checklist Operational tool
Readiness Dashboard System status
Developer Mode Capability-based toggle
RBAC 4 roles, 16 capabilities
Feature Flags 6 flags
Docker Compose Ready for pilot deploy

Next Steps

  1. Deploy pilot environment with Docker Compose
  2. First real upload from phone
  3. First Golden Mission
  4. First week of internal pilot missions
  5. No major architecture changes during first week — only bugs and improvements from real usage

Engineering Standard v1.0

File: docs/ENGINEERING_STANDARD_v1.0.md

  • Grundprincip: Domänen äger sanningen
  • Teknisk stack: React/TS/Vite, Node/TS/Express, Python/PyTorch, Docker
  • Kodstandard: TypeScript strict, ESLint, Prettier, inga any, inga console.log i prod
  • Git-flöde: Issue → Branch → Code → Tests → Commit → PR → Review → Merge → Deploy
  • Kubernetes-first för plattform, GitOps, aldrig manuella ändringar
  • All infrastruktur är kod — samma Git-flöde som applikationskod
  • AI-agent-regler: Arbeta endast i Git, aldrig produktion, skriv tester

LandveX Internal Pilot: DEPLOYED

Status:

Go Live Checklist: docs/GO_LIVE_CHECKLIST.md

Product Levels

File: docs/LANDVEX_PRODUCT_LEVELS.md

  • Level 0: Public (gratis) — öppen karta, trender, heatmaps
  • Level 1: Professional — egna områden, dashboard, rapporter
  • Level 2: Enterprise — AI-regler, Mission Engine, Hotspots, Credits
  • Level 3: Platform — multi-org, egna modeller, white-label, federation

Progressive Disclosure: Grundinställt väldigt enkelt, men man kan gå djupt.

Vision v2.0: Living Operational Model

File: docs/LANDVEX_VISION_v2.md

LandveX är en kontinuerligt uppdaterad operativ modell av kundens infrastruktur som kombinerar verifierade observationer, historik och beslutsstöd för att hjälpa organisationer prioritera rätt åtgärder vid rätt tidpunkt.

Fem nivåer: Reality → Digital Representation → Current State → Intelligence → Prediction

Spatial Intelligence

File: docs/SPATIAL_INTELLIGENCE.md

Tre dimensioner för varje Observation:

  • Semantisk: Vad är objektet? (spricka, skylt, brunn)
  • Spatial: Exakt var? (fasad, våning, zon, höjd, fil, riktning)
  • Temporal: När observerad och hur förändrad?

Precision i 4 steg: GPS → triangulering → 3D-rekonstruktion → historik

OR-001: Operational Readiness

File: docs/OR-001-OPERATIONAL_READINESS.md

  • Every pilot creates assets — Session, Mission, Artifacts, Metadata, Timeline, Report
  • Every failure is a Field Discovery (FD-XXXX) — not a bug
  • Every upload becomes permanent knowledge — Asset → Metadata → Knowledge → Decision → Learning
  • Measure the factory — Reality, Knowledge, Decisions, Learning, Economy
  • Verified Decision Library — biggest asset
  • Sprint planning — starts with real pilot observations

Commits This Session (Full List)

  • e2e3d009 — MASTER PROMPT: Intelligence Lab Development Mode v1.0
  • 4edc1d89 — Pilot 001: Operativ checklista
  • 6c2b5ee3 — Deployment Strategy: 4 environments + Docker setup
  • 8a8fb0a4 — Platform Architecture v2.0: One Platform, Multiple Roles
  • c5a42506 — Readiness Dashboard
  • 207185e5 — Minimal RBAC + Feature Flags + Developer Mode
  • 7b7660b0 — Engineering Standard v1.0
  • 740da921 — Engineering Standard v1.0: Kubernetes-first + GitOps
  • ddfe99f9 — Go Live Checklist + Version Endpoint
  • 6e1aa1b0 — LandveX Internal Pilot: DEPLOYED
  • 1b16e422 — LandveX Product Levels: 4 tiers with Progressive Disclosure
  • 167def5a — LandveX Vision v2.0: Living Operational Model
  • 9883b2c7 — Spatial Intelligence: Three dimensions for every observation
  • 7ceb2b24 — OR-001: Operational Readiness

Stoppregel

Ingen ny arkitektur eller ADR-dokument förrän Pilot 001 genomfört med verkligt uppdrag.

Sprint 0 Mål

En pilotanvändare får en länk, installerar appen via TestFlight eller Google Play Internal Testing, loggar in och genomför ett uppdrag mot https://pilot.landvex.com utan hjälp från en utvecklare.